An essential tool for the care of the diabetic patient is the measurement of blood glucose. Recently, the NIDDK (National Institutes of Diabetes and Digestion and Kidney Diseases) has released the results of a large clinical trial, the DCCT (Diabetes Control and Complications Trial) that shows conclusively that improved blood glucose control reduces the risk of long term complications of diabetes. See, DCCT Research Group, N. Engl. J. Med. 329:977-986 (1993).
Current technology requires that a blood sample be obtained for measurement of blood glucose levels. Samples of venous blood can be obtained from the patient for this measurement, but this method is limited to only a few samples per day, and is not practical for the care of outpatients.
Self monitoring of capillary blood glucose is practical, but still requires multiple and frequent skin punctures. Consequently, most patients perform 2-6 tests per day depending on their personal circumstances and medical condition. Self monitoring results are influenced by technique errors, variability of sample volume and impaired motor skills (important with hypoglycemic episodes). The patient must interrupt other activities to perform the task of blood glucose measurement.
The concept of an implantable sensor to continuously measure the glucose levels in holter monitor type applications and in ambulatory diabetic individuals has existed for several decades. For a recent discussion, see Reach, et al., Anal. Chem. 64:381-386 (1992). The primary focus has been to overcome the disadvantages of capillary blood glucose self-monitoring by developing a glucose sensor, which at the very least, would provide more frequent and easily acquired glucose information. In addition, the sensor could function as a hypoglycemic and hyperglycemic alarm, and ultimately serve as the controller for an artificial endocrine pancreas. The potential limitations of this approach include the limited life of the enzyme, glucose oxidase, the limited lifetime of the sensor (2-3 days), and the need to wear the device.
The concept of a non-invasive glucose sensor has received significant media and technical attention over the past several years. The basic scientific goal has been to utilize near infrared (NIR) spectrophotometry to detect the absorbance properties of the glucose molecule. The inherent problem with this approach is that the glucose signal is weak and is masked by other body constituents. Moreover, if it is possible to detect glucose, the system will most likely rely upon expensive optics and significant computing power, resulting in a large, expensive device which requires frequent recalibration to the patient and provides intermittent data.
Some of the approaches to non-invasive blood glucose measurement are described in U.S. Pat. Nos. 4,428,366, 4,655,225, 4,805,623, 4,875,486, 4,882,492, 5,028,787, 5,054,487, 5,070,874, 5,077,476, 5,086,229, and 5,112,124, the disclosures of each being incorporated herein by reference.
Most of these approaches involve the use of transdermal infrared or near infrared radiation in either a transmission or reflectant mode. In spite of the large number of patents and intense efforts by at least thirty major companies, no devices have been successfully implemented in the field.
The problems with these approaches are well known and described in detail by Marquardt, et al., Anal. Chem., 65:3271 (1993) and Arnold, et al., Anal. Chem., 62:1457 (1990). Marquardt, et al. have shown that in a simple aqueous solution, the absorbance of a 13 mM glucose solution (234 mg/dl) gave a signal with a SIN ratio of about 2. In a protein containing matrix, the actual signal from glucose cannot be detected without considerable manipulation of the data using a partial least squares approach. Such small signal to noise ratios are not practical for developing robust simple instrumentation. Furthermore, the device used in this research is a large spectrophotometer that must be able to scan over reasonably broad wavelength ranges.
In contrast to these purely non-invasive optical approaches, an implant containing a transducer chemical whose optical properties are strongly modulated by recognition of the target analyte will result in a large amplification of the optical signal. It is in this sense that the term "chemical amplification" is used throughout this application. For instance, U.S. Pat. No. 4,401,122 describes an implanted enzymatic sensor that measures the H.sub.2 O.sub.2 produced when glucose and oxygen react in the presence of the enzyme glucose oxidase. This approach is limited by profound biocompatibility concerns, particularly changes in stability related to glucose diffusion to the sensor and the lifetime of an enzyme in an implanted environment. Further concerns using enzymes are created because the large differential between O.sub.2 and glucose concentrations in the body requires a glucose limiting outer membrane. This membrane limits not only the glucose, but the analytical signal as well.
One approach to solving the problems is described in U.S. Pat. No. 5,342,789. In this approach, a fluorescent labeled glycoprotein competes with glucose for binding to a differently fluorescent labeled lectin. Because there is some resonance energy transfer from one label to the other, the presence of glucose reduces the fluorescence intensity of the system. There are two major drawbacks to the system as described in the '789 patent. The first problem is that both labels are photoexcited by the same source; the background signal is significant. The second problem is related to the ability of the system to be implanted into the body. The resonance energy transfer requires diffusion of glucose to the lectin and diffusion of the labeled glycoprotein away from the lectin. In order for the system to have a reasonable time constant for physiological applications, the reagents must be in solution and free to diffuse via a concentration gradient. This makes the device difficult to implement reliably since a reservoir must be designed which allows glucose to diffuse in but prevents the proteins and lectin from diffusing out.
Accordingly, there has been a need for a glucose sensor able to measure glucose over the entire physiological range of 30 to 500+ mg/dl (1.6 to 28+mM). It should provide continuous glucose information and be easy to use. The sensor would not require a sample of blood and would be pain free. From an analytical chemistry standpoint, both the accuracy and the precision would be greater than 95% and the sensor should be non-invasive or minimally invasive. From an instrumental point of view, the device should have a linear dynamic range of at least 200 and a signal to noise ratio of at least 50. Attainment of these figures will ensure that analytical precision and accuracy can be achieved. However, less sensitive instruments could be useful providing measurement of the analyte signals is accurate. The present invention fulfills these needs and provides other related advantages.